DNA sequencing and gene-expression profiling assists in making a tissue of origin diagnosis in cancer of unknown primary

Cancer of unknown primary (CUP) constitutes a group of metastatic cancers in which standardized clinical investigations fail to identify a tissue of origin (TOO). Gene-expression profiling (GEP) has been used to resolve TOO, and DNA sequencing to identify potential targeted treatments; however, these methods have not been widely applied together in CUP patients. To assess the diagnostic utility of DNA and RNA tests for TOO classification, we applied GEP classification and/or gene-panel DNA sequencing in 215 CUP patients. Based on a retrospective review of pathology reports and clinical data, 77% of the cohort were deemed True-CUPs (T-CUP). Among the remaining cases, a latent primary diagnosis (10%) (LP-CUP) or TOO was highly suspected based on combined clinicopathological data (13%) (histology-resolved CUP, HR-CUP). High-medium confidence GEP classifications were made for 80% of LP/HR-CUPs, and these classifications were consistent with a pathologist-assigned diagnosis in 94% of cases, while only 56% of T-CUPs had high-medium confidence predictions. The frequency of somatic mutations in cancer genes was similar to 2,785 CUPs from AACR GENIE Project. DNA features, GEP classification, and oncovirus detection assisted making a TOO diagnosis in 37% of T-CUPs. Gene mutations and mutational signatures of diagnostic utility were found in 31% T-CUPs. GEP classification was useful in 13% of cases and viral detection in 4%. Among resolved T-CUPs, lung and biliary were the most frequently identified cancer types, while kidney cancer represented another minor subset. Multivariate survival analysis showed that unresolved T-CUPs had poorer overall survival than LP/HR-CUPs (Hazard ratio=1.9, 95% CI 1.1 - 3.2, p=0.016), while the risk of death was similar in genomically-resolved T-CUPs and LP/HR-CUPs. In conclusion, combining DNA and RNA profiling with clinicopathological data supported a putative TOO diagnosis in over a third of T-CUPs. DNA sequencing benefited T-CUP tumors with atypical transcriptional patterns that hindered reliable GEP classification.